theorb is now known as theorbtwo
what do you guys think of getting a few of these for a manual tool changer? http://cgi.ebay.com.au/ER32-MT3-M12-COLLET-CHUCK-KEY-CNC-MILLING-LATHE-A75-/150464983915?cmd=ViewItem&pt=LH_DefaultDomain_15&hash=item230869736b#ht_4021wt_1136
i like ER32, but never used it on a mt3 should be good
Valen: you know you want this http://hackadaycom.files.wordpress.com/2008/11/cnc.jpg
Valen: duct tape cnc, no need for your fancy shmancy collets
might have a bit ofrunout
add more duct tape
I think we already have one, it seems to work ok
seems kind of big for MT3. can you get along with a smaller ER?
possibly, I dunno what we already have though
whats the downside to it being big?
(running the mill is more dads thing, I do the computer side of things)
heh just replied to your email about PID tuning btw ;-> (i'm jake)
bigger collets cost more, more Z overhang
Valen: I think most people who responded either didn't read your problem description or didn't understand it
of course on the other hand it could be that I didn't understand it
I think the real win of MT-to-ER is you wouldn't have to mess with the MT drawbar and locking taper
just use two wrenches while the adapter is in the machine to loosen the collet and change the tool
not repeatable tool lengths that way, but neither is MT very repeatable I bet
actually we are planning on getting a few of these and swap the whole thing over
cos its easier than changing the tool lol
just read your response - ouch, not sure how to say it - if you don't know enough hal to do what I said, you might not want to deal with trying such a complex experimental setup. I don't really want to handhold you through it - it seems like it would be a real pain to do remotely.
I've been playing with the shallow end of the pool for a while
I figure HAL isn't impossible to learn lol
would be handy to have a "working model" though
clearly it's not
that's just the same as you wish someone else would do the work for you first
you currently have a working setup - go from that
no problem giving you full access to the machine if you feel like playing
its a fairly large change though from what I understand
it's fairly impossible without being at the machine
If there was an example of this double PID thing that could be adapted or worked from it'd be handy
well I didn't suggest double pid (except as plan-C afterthought)
did you understand what I wrote? maybe I can help if not.
i'm re reading it a sentance at a time
so currently EMC is generating its own velocity number from the difference in encoder positions rather than using mesa's velocity?
ok first up, whats the advantage of a velocity mode control vs torque mode vs "power" which is what I presume the mesa is nativley putting out?
velocity mode hides all nonlinearity in the system from the pid
it also hides disturbances like cutting force
all that happens in the amp
is that a good thing?
I spose PID doesn't handle non linear stuff very well
well it gives better control
what do you mean by better though?
good question. this is all handwavey and I am not a control system expert
hey I don't even play one on TV so your one up on me
I think I have a gut feel for why your FF1 doesn't work and it's because the velocity you get at the axis is not even remotely proportional to the pwm value you give the H bridge
I'd love to just let a neural net run the damn thing lol, let it learn the mill for an hour or 2 ;->
no, the PWM out would vary greatly from velocity due to any number of reasons
but I think maybe you can get real velocity fairly proportional to pid output if you use pid output minus actual velocity to drive the pwm
imagine the axis is moving happily along and you push on it to slow it down. feedback velocity goes down, command-feedback goes up, therefore pwm goes up, pid doesn't have to do all the work
how do you controll the PWM though?
that's what my message is about
thats what I'm trying to get
use a sum2 block as a differential amplifier
er sum2 component it's called (block is old hal terminology)
so drive the PWM as the difference between commanded and actual velocity?
commanded velocity is your pid output
this is (roughly) how velocity mode servo amps work
It feels like the dual PID would give better control over it?, what you describe sounds like a PID with just the P?
PID + P
it feels like PID isn't well suited to things with nonlinear responses
I think that's true
P gain sure isn't, and FF1 isn't
if FF1 had an exponent I could see it working
I spose thats what your recreating somewhat with the dual PID
I gain can cover up a lot of nonlinearity
yeah, it feels all a bit late though
I think you're getting a feel for it. now go experiment!
I've been playing with it for ages lol
we've made a number of things i'm just trying to "tweak it out"
I dont see that FF1 should be anything but linear since it just compensates for BEMF
pcw_home: what about friction force that varies with velocity?
pcw_home: am I talking out my ass? :-)
well if its liear as well it just a little more FF1
the problem I'm having is i can tune the mill during say a G1 F100 to sit at around 0 error
May we all are but ive done 100s of tunings with our PWM amps
so i'll be hovering around +.01 when +ve and -.01 when -ve
then i'll push it to hover around 0.000 with ff1
but if I then change to say a G1 F200 i get an offset again
which gets bigger at F300
I dont recall if it was proportional but i *think* it was
what do you have for an I term
I can "fix" it but then I have issues with overrun and delays with starting
yeah you'll get overshoot with I -- but maybe you can fix it with D
they all add up and swirl together
much more D makes it very noisy
and overshoot is bad with the linear scale system
noisy when sitting?
you can hear it overshoot and correcting at the servo rate and the ferror trace gets noiser
we have wound the preload up so the backlash is really really small, but you can see the motors getting a little inertia up as they wind through it
once its moving though its ok
Better velocity estimation may help the noisiness some
yeah we have a low latency so I'll up the servo rate some
someone should hack our pid to accept that - we've talked about it for years
though I think we are running at 4khz already I think
to handle backlash between motor and encoder?
maybe the linear scale backlash is preventing you from setting P and D high enough
it seems to get a pretty textbook oscilation as P goes up
(that would explain the inability to use I)
I can use I but it seems that even in the textbook world it causes overshoot?
faster servo cycle gives you more ddt velocity noise, not less
the backlash is on the order of .001-003mm or so, our scales measure .001
our P is ~3, D .01 FF1 .01 at F100 to .02 at F200, FF2 .00015 or so
the motors sound "noisy" at a D of much higher than that and it doesn't seem to help things much
You should be able to bump up P if D is higher
It doesn't seem to make much of a difference though
it would reduce the error but wouldn't actually correct it is my understanding?
I have run I up to 50 and see diminishing returns after that, but I've had it up to 200
well you reduce the error to the point the I term can fix the rest
3 sounds low for P, what part of a motor turn gives full scale PWM?
I dont understand the question?
it seems if I could multiply some constant with speed as an exponent and add that to the PID it'd work
Only in the vacuum of space and time
if you disconnect the amp from the motor how far do you have to turn the motor shaft to get full PWM?
how far do we have to move the motor to get the PID to go full sclae
i dont know the answer to that
P doing all the work means its just going to be error * 3 or so isnt it?
when error gets to 1 that would be full scale?
depends on how your pwmgen is scaled
max output is set to .5 I think
is that the right knob?
.001mm per unit
sorry I mean hm2_xxxx.0.pwmgen.xx.scale
I dont know it'll be whatever the default is
I don't have access to the machine running it atm
there are so many scaling factors it's hard to tell how you're set up
surely you don't have max duty cycle .5?
I doubt it
I did change max output to 1 and it made no difference I could see
how did you set P?
winding it up till it oscilated
backing it off a little
adding some D
on our Z drive P is .8 or so
but Z is rather crappy
whats behind your question about movement pcw_home?
Did you try adding more P after the D was added (P will oscillate at very low values without D)
I will give it a go
brb mothers over
I wonder if theres enough spring/hysteresis between the motor and the linear scale to make tuning difficult
I cant see adding more P helping much
other than the offset its pretty nice really
we can see the backlash pretty clearly, Its only really an issue when we change direction
I'll try getting some screen shots
(in the next few days)
If you double P you will roughly half your errors
thats a *lot* of P though
it oscillates when P goes up by ~.5 or so
unless you go with *lots* of D but then its not "smooth"
anybody know about electropolishing?
WalterN, you need the right mix of chemicals for your material, so you need to find docs from a chemical supplier to match the material
archivist, 316 stainless?
[07:40:29] <archivist> http://www.google.co.uk/#hl=en&source=hp&q=electropolishing+stainless+steel&aq=0&aqi=g10&aql=&oq=electropolishing+&gs_rfai=&fp=63b42caf01504688
is low voltage typically used?
or does total watts matter the most?
did you read the page :)
bleh, not really :P
never mind the volts feel the amps
read that page you linked
"Electrolytes used are usually mixtures of phosphoric and sulphuric acids."
would just sulfuric acid work?
its the easiest to get...
you probably need a mix depends
how about nitric acid?
maybe throw in some sulfuric too
I ask because I have a gallon of parkerizing stuff coming, evidently all that is is nitric and phosphoric acid
and some salt..
nvm, probably not a good idea
with that stuff anyway
archivist, so does the current and volt matter more than just measuring in watts?
yes its amps per area
like, could I do 170volts and whatever amps?
no its around 30 amps per 10 cm square
so its the amps that matter the most, not the volts
and whatever the wattage turns out to be
3amps per square cm?
with the right chemicals and you achieve the voltage mentioned you should be running at the correct amperage
I say 170 because thats just 120volts turned into DC current, which would be about 170v
you vary the voltage to achieve the current density
massive tansformer needed
I was kinda hoping to skip such a transformer
or series of tansformers
is there a good place to get a massive transformer like that?
normal transformers wont cut it
[08:11:52] <WalterN> https://www.surpluscenter.com/item.asp?item=11-2840&catname=electric
you get them wound for you
theres got to be off the shelf stuff though...
CNC machines have big arse transformers in them..
I doubt it
archivist, well, thanks.. got some good info anyway
bricofoy is now known as bricografcetmanf
* bricografcetmanf finnaly got the grafcet working !
bricografcetmanf is now known as bricografcetman
how can I contribute to improve the doc on grafcet ?
It miss just one or two little things to make it understandable and usable, and if Chris Morley had not pointed out his grafcet sample, I woudl have given up the thing
bricografcetman is now known as bricofoy
made a coloring rule for gedit, for HAL files
looks like this : http://imagebin.ca/view/Hnbt3c7.html
if someone is interested, the file is here : http://filebin.ca/wjhscv/hal.lang
this file must be located in /usr/share/gtksourceview-2.0/language-specs/hal.lang, then restart gedit
it is based upon an esisting coloring rule for another language, so it do not fully work as I expected, but still give some eye-candy to boring hal files :)
not perfect, but for me its enough. but feel free to improve it :)
planned to do the same for GCode files, but found that the existing "Haskell" rule give interesting results. Again, enough for me :) I think just copying the haskell rule into a gcode.lang new file and changing the magic to reconize gcode file will be enough
here is what gcode looks like with "haskell" rule : http://imagebin.ca/view/p5HHQg.html
power went out... all from a little rain! very annoying.
glad the computer is on a UPS :)
router and cable modem, not.
elmo40_ is now known as elmo40
only took about 3 hours to find the bug which was causing drop-cutter slowness....
oh, #emc not #cam... :|
elmo40, my two computers, router, and modem thing are hooked up to my UPS
actually, my modem is and the main router, but my computer is on the wifi router... i should change that ;)
I have gig ethernet router between my two computers
that's a waste
why would you wnat a ROUTER instead of a SWITCH
because it also has dual band wireless for other things that connect to it
like laptop and cellphone
Oh, you're using a consumer router/wifi thingy.
its good enough for the hobby server I have set up too
at least with dd-wrt installed on it
hi i'am on xbuntu RTC hardy can i get there gedit as a default instad of mousepad ?
EMC is running
yeah ! my grafcet works ! (after a full week fighting with the bugs of the editor, and turning around the bugs of the interpreter :)
ries_ is now known as ries
[#emc] "This is the #emc channel - talk related to the Enhanced Machine Controller and general machining. Website: http://www.linuxcnc.org/,
wiki at http://wiki.linuxcnc.org/"
is it possible to use emc2-sim in order to check a g-code files machining-time in advance?
I think so
depends how complicated the g-code is
does it rely on inputs from the outside, tool-changes, etc
loops based on measured stuff
no, the only thing making it difficult to estimat without a simulation is the acceleration
if you set up your accel and velocity to match your real machine, the simulation will be really close
thats wahat i am looking for, cradek
if it's straightforward, you can run in sim and have it close
is there a speedup-mode for the simulation?
you can set your accel, vel, and feed override as high as you want
so I will give it a try...
there is no setting in sim that can speed things up past the accel and vel limits
so you can't "accurately" simulate the amount of time a file would take to machine and also have a high speed simulation using the same config
no, those are contrary goals
note also that the AXIS time estimate doesn't take accel into account
yes, i already figured that :(
thats the reason why i am going to try the simulation
I wonder if it takes max velocity into account
I believe it just uses the programmed feed rate for each move, even if it's beyond the machine's capabilities
what does "Invalid configuration of axes is preventing EMC from starting" mean?
EMC v2.3.5 runs fine, but upgrading to 2.4.2 I get that error
it means there's a different error higher up, and you didn't read the UPDATING wiki page
ah, thank you.
there's an ini file change required for 2.3->2.4 upgrade
is that information also in a README that gets installed with the package?
(or some other file)
I don't know
I think it should be, I guess :)
I can sympathize with that opinion, but I also know nobody would read it there
the instructions are in the release announcement emails and the first link on the main page of the wiki...
I guess when people ask the same question over and over, it means we should have done something differently
well, or they should have (read the release notes for instance)
that's probably why I thought of it :)
I wonder if there is any way to tell if a package is being installed vs upgraded
(at install time, so that some postinst triggers can be changed or something)
make the run script check the VERSION in the ini file, and less README if it doesn't match :)
SWPadnos: there is - when dpkg installs a package, it runs the debian "maintainer scripts" such that they can tell if they're being upgraded, and if so from what version to what version
(or gedit README, since it may be run from an icon)
I don't know if that actually helps though, since an update/upgrade could be part of a large set of packages, and we wouldn't want to hold up the entire works to display a text file
lots of packages do that
maybe it's ok if we do too, at least when doing upgrades that require users to change their config files
another option is to stick an extra document in the CNC menu if it's an upgrade
anyone else get a "PARPORT: linux parport parport0 does not support mode 4." after upgrading?
what parport-using hal driver?
EMC v2.4.2 Hm2 v0.15, 7i43 v0.3
looks like it still works ok.... I double checked in the bios. only one PP, and it's set to EPP
dmesg -> http://pastebin.com/v36qP1Au
concrete 3d printer: http://blog.makezine.com/archive/2010/07/cement_printer_makes_large_3d_sculp.html
eeh. lookit the spindle on that http://cgi.ebay.co.uk/SCM-Routron-CNC-Router-921-/250658924168
You know, I haven't seen any total noobs turn up here for a while. Is that because I have been to busy actually making stuff, or is there a bit of a drought?
"discretionary reserve of 500 squids"
Yiou buying it?
what about all the ones I've been sending from #reprap? :)
Perhaps they go to bed early :-)
nope not buying anything atm. I can't even afford new brushes for mine
I'm hoping they last a few more days until these mendels are made
I wish I was creditworthy
I managed to break the screw-on-nose screw off of my Dremel-a-like.
Hi Andy, we will get you your 8I20s but we're still mucking about with the host side firmware
an overdraft facility would be reet gurt right now
Pretend you live in Botswana and apply to Kiva?
PCW: I need to get that 5i20 order placed too.
The host side interface will fit in a 5I20 (just, 6 axis sserial + 6 encoders = 99% full)
Looking elsewhere on the interwebs that Routron is metal. The one in the eBay ad looks to be wood.
Is the 400kgate 7i39 actually more roomy then?
seb_kuzminsky: nice machine, but they are using it for sculptures of poop :)
Yes, 5I20 is 200K "xilinx gates"
I am considering building an Atom system with more RAM than HD.
(Well, the same amount, 4GB SSD HD and 4GB or ram)
I'd still like to see a raid of SD done
Like this? http://www.youtube.com/watch?v=96dWOEa4Djs
not quite :) I was thinking 5 or 10 cheap SD cards to replace one SSD
those 32GB ones are pretty cheap now. the ones with fast read but slower write, just for booting
$79 aka 100 quid + vat
32GB ssd, not sd
I am using an 8GB DOM for the Miller controller.
I spent an evening trying to work out if I could get a decent IDE ssd or hdd for my old thinkpad, but in the end just went to Wilko's and bought a usb stick. job done
Mainly for the simple mounting.
How to get an OS + EMC onto a machine with no CD-rom drive, I wonder?
usb cd rom?
coreboot? I have an eprom programmer now, but haven't used it for anything yet
PXE is a possibility.
Secondary problem is that is has no monitor yet. I guess SSH into a machine with no OS is going to be tricksy :-)
atom boards generally have vga don't they?
it's compulsory, as you have to use the intel glue chip
eeh, glue chip :) it's like the good old days again :)
PCs have finally caught up with spectrums